Skip to content

Zero-inflated model inquiry

2 messages · Peter Houk, Drew Tyre

#
Hi Peter,

Your assumption that Before and During are contrasted with After is
correct. By default R parameterizes categorical variables using
treatment contrasts which compare each level to the first one, and the
default sorting is lexicographic, so AFTER becomes the first level.
Your model is indicating that the average abundance both BEFORE and
DURING are significantly different from the AFTER. It sounds like what
you'd like to know is also BEFORE different from DURING. I see a
couple things you could try
1) Make predictions of the average urchin_abundance from the model for
each period along with confidence intervals. Use the confidence
intervals to decide what is the same and different.
2) Change your formula to urchin_density~impact_period-1. This will
give you a distinct estimate for each period, and make construction of
the confidence intervals in 1 very easy, but still won't give you all
the pairwise comparisons.
3) Check the package multcomp and use it to find the appropriate
contrasts for all three levels. I'm not sure this will work for models
from the pscl package.

hth
On Tue, Sep 25, 2012 at 10:50 PM, Peter Houk <peterhouk at gmail.com> wrote: